NNPDF / nnusf

An open source machine learning framework that provides predictions for all-energy neutrino structure functions.
https://nnpdf.github.io/nnusf/
GNU General Public License v3.0
0 stars 0 forks source link

Dump NN SF predictions into LHAPDF grid #59

Closed Radonirinaunimi closed 2 years ago

Radonirinaunimi commented 2 years ago

Add modules that dumps the NN predictions into LHAPDF-like grids.

For the time being, the IDs are defined as below: SF ID
$F^{\nu}_{2}$ 1001
$F^{\nu}_{L}$ 1002
$x F^{\nu}_{3}$ 1003
$F^{\bar{\nu}}_{2}$ 2001
$F^{\bar{\nu}}_{L}$ 2002
$x F^{\bar{\nu}}_{3}$ 2003
$\langle F_2 \rangle$ 3001
$\langle F_L \rangle$ 3002
$\langle x F_3 \rangle$ 3003
juanrojochacon commented 2 years ago

F3 means xF3 right?

Radonirinaunimi commented 2 years ago

F3 means xF3 right?

Yes, indeed! I made this explicit in the description above.

juanrojochacon commented 2 years ago

Also these are for proton SFs? Do we also have grids for other nuclei?

Radonirinaunimi commented 2 years ago

Also these are for proton SFs? Do we also have grids for other nuclei?

These are only for proton SFs (sorry, I could have made this explicit in the naming), I will generate the other nuclei now.

Radonirinaunimi commented 2 years ago

This is now ready to be merged if anyone has time to review it. Otherwise I'd also be happy to merge it.

RoyStegeman commented 2 years ago

Since it performs a very simple task and doesn't really touch any existing code I would be happy to merge it without review.

Though from a code perspective i"m fine with this, I don't think we should publish the average in this way as mentioned on the slack channel.

Radonirinaunimi commented 2 years ago

Since it performs a very simple task and doesn't really touch any existing code I would be happy to merge it without review.

This is indeed something non-controversial but really important to get bug-free. So far, the results suggest that this is the case, and some of the functions anyway I took and modified from EKO. I am indeed in favor of merging this asap (without review).

Though from a code perspective i"m fine with this, I don't think we should publish the average in this way as mentioned on the slack channel.

I am not sure I understand what you mean about the average? I couldn't find a reference of this on the slack.

RoyStegeman commented 2 years ago

Nvm, forget I said anything. Feel free to merge